【作者】 宗欣露；

【导师】 熊盛武； 方志祥；

【作者基本信息】 武汉理工大学 ， 计算机应用技术， 2011， 博士

【摘要】 人车混合疏散是目前现代交通学科、对地观测学科、地理学科、人工智能学科、计算机学科、现代通信学科、以及公共卫生与安全等领域所共同面临的一个重要研究问题与急迫课题。目前国内外的疏散理论研究不能满足人车混合疏散的研究需求,其中疏散模型基本都集中于人员的疏散,交通流模型则直接面向车辆,而对于我国人车混合通行的交通特点,现有疏散研究缺少人车混合疏散运动模式的分析、人车混合疏散行为的演化规律的研究,从而在现实中缺少适用性,也无法形成人车混合的科学疏导决策。另一方面,目前从优化角度进行疏散建模的研究不多,已有的优化模型也大多是考虑单一交通模式在一个目标上的优化或者将多个目标转化为单目标优化,其得到的单个最优解无法为管理者提供多个目标上的决策依据。鉴于此,本文依据计算机仿真与多目标优化相结合的研究思路,为解决上述问题展开了系列研究：提出了大型活动场所人员疏散模型和基于蚁群算法的层次多目标疏散路径算法。针对大型场所多目标应急疏散路径问题,构建了层次引导网络,描述了一种面向目的的组织结构,在寻找出口的过程中为人员提供多级指导,只有当处于某个层次网络中的人员到达相邻的下一层次网络时,才允许其进入下一层网络直至最终到达出口所在的层次,该层次网络能够有效避免人员疏散过程中的盲目性,在层次引导网络的基础上,提出了一种多目标优化疏散模型,同时优化三个疏散目标,分别为最小化疏散时间、疏散距离和拥堵程度；提出了基于蚁群优化算法的层次多目标疏散路径算法,应用于大型体育馆内部人员疏散问题。数值实验结果表明,与K最短路径、NSAG-Ⅱ、基本蚁群算法相比,多目标疏散模型和基于蚁群算法的层次多目标疏散路径算法能够为大型建筑物内部人员疏散问题提供多个高效、安全的时空路径疏散方案。提出了基于蚁群优化算法的人车混合疏散模型。在分析了人车混合通行特征和混合疏散目标体系的基础上,针对突发事件下的人车混合疏散问题,以人车混合疏散的总时间最短、混合道路利用程度最高为目标,建立了一种人车混合疏散的多目标优化模型,设计了蚁群求解算法,并提出了带有自适应禁忌调整和限制阈值的信息素更新策略的多目标蚁群优化算法。通过应用于大型体育场及其周边路网集成环境,实验结果表明该模型及算法对人车混合交通流疏散问题具有良好的效果,尤其是当行人所占比例为50%——80%时,人车混合疏散效果在两个目标上较优,可以为大型场所的安全出行和突发事件下的人车混合疏散方案的制定提供一定的理论指导。提出了基于多蚁群系统的人车混合疏散模型。在研究了人车混合疏散过程中的行为特征以及相同对象之间、不同对象之间的相互影响的基础上,建立了总疏散时间最小、整个路网交通负载均衡的疏散模型,针对单一蚂蚁系统的正反馈机制可能导致某些较优路径上的拥堵这一不足,根据疏散过程中的群体效应提出了基于多蚁群系统协同进化的方法来解决人车混合疏散问题,用多个蚁群系统模拟疏散过程中不同的交通对象之间的竞争和影响,利用蚁群间的通信机制模拟人员和车辆的交互。数值实验结果证明基于多蚁群系统的疏散模型和算法在Pareto解的分布、疏散效率和个体出口分布等方面均优于基于单一蚁群系统的方法,体现了多蚁群协同进化算法在解决人车混合疏散问题上的优越性。提出了基于粒子群的人车混合疏散模型。针对疏散个体在紧急情况下的疏散过程中所呈现的一些特殊心理和行为：从众心理和行为、小群体现象,利用粒子群优化理论来模拟和优化人车混合疏散过程,定义了人车时空冲突和时空拥挤度的概念,建立了基于时空冲突和时空拥挤度最小为目标的疏散模型,提出了带邻域学习因子的离散粒子群优化算法,将疏散个体看作粒子,粒子的运动除了受到自身经验和群体最优个体的影响之外,同时还向邻域中的最优个体学习,这种局部和全局并存的学习机制既模拟了疏散过程中的从众行为,又加速了寻找安全出口的进程。实验结果表明,基于粒子群的人车混合疏散模型可以应用于人车混合疏散仿真,并在一定程度上优于基于蚁群算法的疏散模型。更多还原

【Abstract】 Pedestrian-vehicle mixed evacuation is an important and urgent issue, which confronted commonly by modern transportation, earth observation, geography, artificial intelligence, computer science, modern communication and public health and security. The existed research on evacuation, on the one hand does not reach the demand of evacuation mixed with pedestrians and vehicles. Most of evacuation models focus on pedestrian evacuation inside building while the models based on traffic flow are directly vehicle-oriented. As for passing characteristics of pedestrian-vehicle mixed traffic in China, however, the current studies on evacuation are lack off analysis on movement pattern and evolutionary process of pedestrian-vehicle mixed evacuation, which leads to inapplicable in practice and difficult to form scientific evacuation measures. On the other hand, there are few works studying evacuation modeling from the perspective of optimization, the existing optimization models only consider optimizing one objective for single transportation mode or converting multi-objectives optimization problem into single objective optimization problem, whose single optimal solution can not offer manager decision foundation on multiple objectives.To solve these problems described above, based on the computer simulation and multi-objective optimization, many researches are carried out:Focusing on the evacuation routing problem in large public place, a hierarchical directed network is presented, which describes a destination-oriented structure. The hierarchical directed network offers hierarchical guidance for pedestrian during the process of searching for exits. Only a pedestrian whose current position belongs to a certain rank achieves the rank’s corresponding destination, he/she is allowed to enter into next rank and finally reach one of the terminal exits. This strategy can effectively avoid blindness in evacuation process. On the basis of hierarchical directed network, a multi-objective optimization evacuation model is proposed to minimize three objectives simultaneously, total evacuation time, total evacuation distance and cumulative congestion degrees. A hierarchical multi-objective evacuation routing problem algorithm based on ant colony optimization algorithm is designed, which is tested using a stadium. The results of four different optimization algorithms are analyzed, and congestion degree, space-time distribution of pedestrian number and evacuation paths are discussed. The proposed model and algorithm in this paper can solve the problem and provide multiple safe, efficient evacuation plans.Based on the analysis of characteristics of traffic flow mixed pedestrians and vehicles, a multi-objective optimization model was proposed to tackle pedestrian-vehicle mixed evacuation problem under emergency situation. This model aims to minimize total evacuation time and maximize mixed road utilization of the whole road network simultaneously. To solve this model, a multi-objective ant colony optimization algorithm is designed. In order to tackle the deficiency of the algorithm in pedestrian-vehicle mixed evacuation problem, an improved ant colony optimization algorithm with proper heuristic information, self-adaptive tabu adjustment and pheromone updating strategy constrained by threshold is described. The proposed model and algorithms are tested using a case integrated a stadium of Wuhan Sports Center (in China) and road network around it. Evacuation performances with different mixed proportion of pedestrians to vehicles are analyzed. The experimental results show that this model and algorithm are effective for evacuation problem with mixed traffic flow, and mixed evacuation results are better when the mixed ratio of pedestrian ranges from 50% to 80%. By comparing the solutions and spatial-temperal performances of two methods, the improved approach is better in solving pedestrian-vehicle mixed evacuation problem. The research in this paper can offer decision support for planning pedestrian-vehicle mixed evacuation in large common place.Studying the behavior characteristics of evacuees, interrelationship between the same individuals and interaction between different individuals in the process of pedestrian-vehicle mixed evacuation, evaluation criteria such as minimal evacuation time and balanced traffic load, are designed. The positive feedback mechanism of single ant colony system may lead to congestion on some optimum routes. Like different ant colony systems in nature, different components of traffic flow compete and interact with each other during evacuation. According to the group effect in evacuation process, an approach based on multi-ant colony system evolution is proposed to tackle mixed traffic flow evacuation problem. A multi-objective model is established to minimize total evacuation time and balance traffic load of the whole road network. Communication mechanism between colonies is used to simulate the interaction between pedestrians and vehicles. Performances of the two approach based on single ant colony system and multiple ant colonies system are analyzed, including distribution of solutions, evacuation efficiency and distribution of individuals in each exit. The experimental results indicate the superiority of coevolution of multi-ant colony system over single ant colony system in mixed evacuation problem.By studying some special psychology and behaviors of individuals in emergency evacuation, including the psychology of going with the crowd, sub-group phenomenon, and swarm intelligence theory is utilized to simulate and optimize evacuation process mixed with pedestrians and vehicles in this paper. Pedestrian-vehicle spatial-temperal conflict and spatial-temperal congestion are defined. And an evacuation model base on minimizing both spatial-temperal conflict and congestion is presented. Discrete particle swarm optimization with neighborhood learning factor algorithm is proposed to solve pedestrian-vehicle mixed evacuation problem. By the approach each evacuation individual is considered as a particle. The particles moves impacted by self-experience and best individual in swarm, in addition, they learn from the best particle in their respective neighborhood particles. This local and global learning mechanism simulates the individual’s behavior of going with the crowd as well as accelerates the process of searching exits. The simulation results of ant colony optimization algorithm, multi-ant colony optimization algorithm, particle swarm optimization algorithm and discrete particle swarm optimization with neighborhood learning factor algorithm are compared and analyzed.更多还原